Fabrication of integrated circuits with feature sizes on the order of nanometers requires etch processes that are extremely uniform across the entire surface of a semiconductor wafer, typically about 300 millimeters in diameter. Such uniform etch processes are typically realized in a plasma etch reactor such as a capacitively coupled plasma reactor with an overhead electrode having an extremely high aspect ratio (e.g., an electrode-to-wafer gap of only 2.5 cm for a 300 mm wafer). The etch process gas includes a gas of the type that forms a protective polymer layer over photoresist or other thin film surfaces that are not to be etched, such as a fluorocarbon gas or a fluorohydrocarbon gas. Formation of such a protective polymer layer enhances etch selectivity.
Such processes exhibit a low etch rate, etch stop or taper profile at a region typically the center of the wafer and a high etch rate or bow profile at other region typically near the wafer periphery. Such a center low etch rate, etch stop or taper profile has appeared to be unavoidable for a number of reasons. First, the process gas is introduced either from the side of the wafer or over the top of the wafer. Evacuation of gas from the reactor chamber requires the gas to flow radially outwardly across the surface of the wafer, so that gases fed to the wafer center pass over the wafer periphery before being evacuated. Therefore, the residency time of the gas increases with wafer radius, so that the minimum residency time (and hence the minimum gas dissociation) occurs at the wafer center. This effect is particularly pronounced due to the high aspect ratio of the reactor chamber. This aspect ratio arises from the small electrode-to-wafer gap (e.g., about 2 cm) and the large wafer diameter (e.g., 300 mm). The low dissociation of plasma species at the wafer center results in more complex (e.g., more carbon-rich) fluorocarbon or fluorohydrocarbon plasma species at the wafer center, which tend to etch dielectric material on the wafer more slowly while depositing etch-stopping polymer more quickly, thereby depressing the etch rate at the wafer center. In contrast, the high dissociation of plasma species at the wafer edge produces simpler (more active) etch species that are relatively high in fluorine content (the extreme example of such reactive species being free fluorine) and produce low-carbon content polymer films that accumulate more slowly on photoresist surfaces. At the wafer center, the effect of low dissociation is most noticeable when etching extremely narrow openings through a dielectric thin film. At the wafer center, the polymer accumulates on the side walls of the narrow openings causing taper profile angle, and eventually, as the opening depth increases, pinches off the opening and stops further etching of the opening before the required depth is reached, a phenomenon referred to as etch stop.
These problems are exacerbated when attempting to increase the etch rate or improve profile by the expedient of increasing the process gas flow rate into the reactor chamber. Such an increase in gas flow rate does not solve the problem of non-uniform residency time radial distribution across the wafer surface (and hence the non-uniform dissociation across the wafer surface), and may even worsen the problem, thereby worsening the etch rate or etch profile non-uniformity. One approach to improving the etch rate or etch profile at the wafer center might seem to be increasing the process gas flow rate over the wafer center or decreasing it at the wafer periphery (or both). However, such a technique would further decrease the gas residency time (and dissociation) over the wafer center relative to the periphery, further worsening the etch rate or etch profile non-uniformity. Thus, there appears to be no solution to the problem.
Another reason for such problems is the process gas content. We have found that the problem arises with great severity when using particular combinations of fluorocarbon or hydrocarbon etch process gases that otherwise produce the best possible process results, such as (for example) a process gas that has, for its etchant component, C4F6 and/or CH2F2. Another example can include CHF3 and/or CF4. These process gases have been found to produce desired results (with the exception of the tendency for etch stop discussed above) when etching such dielectric materials as silicon dioxide or silicon nitride or low dielectric constant materials such as porous organo-silicate glass or nitrogen-doped silicon carbide, for example. It has seemed that the only way of avoiding center-low etch rate distribution or the related taper profile or etch stop problems is to employ other (less desirable) process gas mixtures.
Another cause for the center-low etch rate distribution across the wafer surface arises in a particular type of capacitively coupled etch reactor. In the beginning, a capacitively coupled etch reactor employed a single RF power source coupled between the wafer and the overhead electrode. In such a reactor, the etch rate could only be increased (to enhance productivity) by increasing the RF power. Such an increase unfortunately increases the ion energy, causing more bombardment damage to photoresist and thereby reducing etch selectivity. This problem was circumvented by introducing low frequency magnetic fields at the sides of the chamber (in lieu of increasing the RF power) to improve the etch rate, in which case the reactor is called a magnetically enhanced reactive ion etch (MERIE) reactor. This approach was successful in improving the etch rate (e.g., by a factor of 2 in some cases) without damaging the photoresist or reducing the etch selectivity. It is felt that the MERIE magnetic field boosts the etch rate by increasing ionization. Recently, the RF source power has been decoupled from the ion energy or photoresist bombardment damage by applying VHF source power that contributes primarily (or almost exclusively) to ionization while applying independently a low frequency (or HF frequency) bias power that contributes primarily (or almost exclusively) to ion energy. This permits ionization and/or dissociation to be increased, without increasing ion bombardment damage to photoresist, by increasing the VHF source power without increasing the lower frequency bias power. Nevertheless, even with such a dual frequency approach for decoupling control of dissociation and ion energy, MERIE magnets can be used. The problem is that the MERIE magnetic fields tend to have their greatest etch rate-enhancing effect nearest the wafer edge. This produces a center-low etch rate distribution across the wafer surface, which has seemed to be an unavoidable characteristic of MERIE reactors. Typically, the MERIE reactor also suffers from the effects (discussed above) of low gas residency time over the wafer center, that causes center-low etch rate distribution. The relatively high dissociation achieved in such a reactor, through the use of VHF source power and MERIE magnets, makes the non-uniformity of the dissociation (due to non-uniform gas residency time across the wafer) more critical.